Environ. Sci. Technol. 2003, 37, 5191-5196
Solid-Solution Partitioning of Cd, Cu, Ni, Pb, and Zn in the Organic Horizons of a Forest Soil S EÄ B A S T I E N S A U V EÄ , * , † S Y L V I E M A N N A , ‡ MARIE-CLAUDE TURMEL,‡ A N D R EÄ G . R O Y , ‡ A N D F R A N C¸ O I S C O U R C H E S N E ‡ Department of Chemistry and Department of Geography, Universite´ de Montre´al, P.O. 6128, Succursale Centre-Ville, Montreal, Quebec, Canada H3C 3J7
We report the solid-liquid partitioning of Cd, Cu, Ni, Pb, and Zn in 60 organic horizon samples of forest soils from the Hermine Watershed (St-Hippolyte, PQ, Canada). The mean Kd values are respectively 1132, 966, 802, 3337 and 561. Comparison of those Kd coefficients to published compilation values show that the Kd values are lower in acidic organic soil horizons relative to the overall mean Kd values compiled for mineral soils. But, once normalized to a mean pH of 4.4, the Kd values in organic soil horizons demonstrate the high sorption affinity of organic matter, which is either as good as or up to 30 times higher than mineral soil materials for sorbing trace metals. Regression analysis shows that, within our data set, pH and total metal contents are not consistent predictors of metal partitioning. Indeed, metal sorption by the solid phase must be studied in relation to complexation by dissolved organic ligands, and both processes may sometime counteract one another.
Introduction Solid-solution partitioning is critical to assess the potential leaching of metals and their bioavailability in soils. There is a large amount of soil partitioning data available in the literature (e.g., refs 1 and 2). Much of these data are focused on mineral soils and as a result, there is a paucity of data on organically dominated materials (i.e., LFH horizons or O horizons from organic soils). A standard threshold of 20% carbon by weight is used to separate mineral from organic soil horizons (3). Nutrient cycling and metal leaching in many forests depend on the properties of the surface organic soil horizons where roots, nutrients, and soil biota are concentrated. As such, solid-solution partitioning of trace metals is expected to be very different in organic acidic soils relative to that of agricultural or urban soils inasmuch as pH, texture, organic matter inputs, soil biota, and plant/tree cover will be very different in these environments. It is somewhat surprising that so few data pertain to organic soil horizons, especially considering that it is such an ubiquitous component of forest soils which are themselves widespread. In this context, our data set will be a great help to characterize and model metal partitioning in forest soil horizons that are dominated by organic matter and that represent large areas of the world. * Corresponding author phone: (514)343-6749; fax: (514)343-7586; e-mail:
[email protected]. † Department of Chemistry. ‡ Department of Geography. 10.1021/es030059g CCC: $25.00 Published on Web 10/16/2003
2003 American Chemical Society
Solid-solution partitioning is usually evaluated using the ratio between the total solid concentration of metals (in mg kg-1) over the concentration of solution dissolved metals (mg L-1). The resulting Kd coefficient (eq 1) is therefore usually reported in (L kg-1):
Kd )
Total Metal (Dissolved Metal)
(1)
Some averages reported in the literature for Kd coefficients of mineral soils are Cd ) 3000, Cu ) 5000, Ni ) 17 000, Pb ) 170 000, and Zn ) 12 000 (2). It must be emphasized that Kd coefficient values for a single element may vary over nearly 6 orders of magnitude mainly as a function of soil pH, soil total metal, and soil organic matter contents (2). The relationships between Kd values and physicochemical characteristics of the soil also vary among different trace metals. For instance, most divalent metals have Kd values that are sensitive to pH variations: Cu is reportedly preferentially sorbed to organic materials, and Pb has a high affinity for iron oxides (4). The data and the semiempirical equations relating the Kd values to soil pH and soil organic matter are therefore not necessarily applicable to risk assessment and environmental fate modeling in the organic soil horizons of forested sites. A study was instigated to examine the spatial variability of the properties of organic soil horizons within a small forested watershed using a spatially dense sampling pattern. The field sampling program focused on one hand on the relationships among microtopography, soil chemical parameters, and tree species distribution (5). On the other hand, the data offer an excellent opportunity to quantify the variability of the partitioning of trace metals in organic forest soils. Furthermore, unlike many published partitioning studies, our data set is made up only with naturally occurring metals in a relatively pristine environment, far from sources of trace metals. In this study we did not use metal spikes. Given the relatively large number of samples from the same site, it is also an opportunity to explore the relations among some soil properties within a relatively small range of in-situ values. Our objectives are (i) to measure the Kd coefficients from a data set of 60 samples of soil organic horizons collected in a small forested watershed, (ii) to contrast the resulting Kd coefficients with those reported or predicted using published compilations, and (iii) to identify which soil physicochemical properties control metal partitioning within a data set showing a small spread in some of the controlling variables (e.g., pH).
Experimental Section Sampling Site. The study watershed (Hermine) is located near St-Hippolyte, 80 km north of Montre´al, PQ, Canada (Figure 1). The Hermine is a 5.1-ha catchment drained by an intermittent first-order stream. The vegetation is mainly composed of deciduous species such as sugar maple (Acer saccharum), American beech (Fagus grandifolia), and yellow birch (Betula alleghaniensis). The soil underlying the forest canopy is a Podzol developed from a glacial till and having a mineralogy similar to that of the anorthositic bedrock [mostly quartz with some plagioclase and K-feldspars (6)]. The study site is an acidic forested soil with no measurable atmospheric metal inputs. Given the relatively low density of soil macro-biota in acidic soils, there is little opportunity for the integration of minerals into the upper organic soil horizons. VOL. 37, NO. 22, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5191
FIGURE 1. Location of the experimental site. Sampling Design. Fresh litter (L), fibric (F), and humic (H) horizons were sampled in the fall of 2000 every 10 m along four transects oriented perpendicular to contour lines (see Figure 3 in Supporting Information). The soils were then sieved dry to 99% of soil solution dissolved metals are bound to soluble organic ligands; 13, 14). Therefore, it is possible that the partitioning coefficients of Cd and Ni tend to be similar to that of Cu simply because Cd and Ni have a lower affinity to sorb to soil organic matter concomitant with a lower affinity to form dissolved complexes with organic matter in solution (e.g., refs 4, 15, and 19). The competition of complexation by dissolved organic matter and sorption/precipitation to the solid phase can also explain the apparent higher sorption of Pb. The Kd values for Pb are the highest ones observed in this study (Figure 2). This is partly due to the very low mineral solubility of Pb. Also, as compared to Cu, Pb has a higher tendency to sorb to soil solid organic matter than to complex with dissolved organic matter in solution. While Cu, relative to Pb, has a strong tendency to complex with dissolved organic matter rather than to sorb to the organic matter in the solid phase (Table 1). This result is not too surprising and concurs with the work of Saar and Weber (16), who reported that Pb has a higher affinity than Cu or Cd for coagulating or precipitating fulvic acids and that, conversely, Cu has a higher affinity for the more labile, dissolved fulvic acids. It must also be emphasized that this does not mean that Pb has a lower relative affinity for dissolved organic matter, it only means that it has a much higher affinity for the solid phase (mineral or organic). Comparison with Published Data. We compared our experimentally determined Kd values to those compiled from a literature review (Table 3). The comparison suggests that mean Kd values in the organic soil horizons are about 3-15 times lower than those reported for mineral soils. Given that pH and soil organic matter are critical parameters that influence solid-liquid partitioning, we used some published equations to normalize the effect of soil solution pH (see Table 2 in ref 2). Even though some equations have been developed for the normalization of soil organic matter content, we could hardly use these equations because the 5194
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 22, 2003
TABLE 3. Comparison of Mean Kd Values Obtained for the Organic Horizons from This Study and from Published Values with and without pH Normalizationa element
Kd (organic)b
Kd (compilation)c
Kd (pH 4.4)d
ratio (organic/pH 4.4)
Cd Cu Ni Pb Zn
1132 966 802 3337 561
2 869 4 799 16 761 171 214 11 615
36 476 26 3357 57
31.5 2.0 30.6 1.0 9.8
a The ratio of the K values experimentally determined in this study d is reported relative to the partitioning coefficients expected in mineral soils normalized to the mean pH of our data set (4.4). The ratios are also calculated to facilitate the comparison. a Experimental data for organic horizons from the current study. c Compiled data for mineral soils (2). d Compiled data for mineral soils normalized to a mean pH of 4.4 (2).
organic soil horizons in our study have organic matter contents that are well outside the range of applicability of these semiempirical regression equations (2). After normalization of the compiled mineral soil Kd values to a mean pH of 4.4, the partitioning coefficients of our data set are no longer lower than those reported for mineral soils. In fact, the mean Kd values we have measured here are always equal to or larger than the normalized Kd values of mineral soils (almost identical in the case of Pb, twice higher for Cu, 10 times higher for Zn, and about 30 times higher for Cd and Nissee calculated ratios in Table 3). This is in agreement with the known high sorption affinity of organic matter for metals. It also shows that pH normalization is critical for a proper evaluation of metal sorption. The variations in the ratios of Kd values for organic versus mineral materials are reported in Table 3. Given that the Kd values for organic soil horizons are similar across the five metals (Table 2 and Supporting Information Figure 3), the variations in the organic to mineral ratios can be interpreted as a reflection of the variation in the affinity between metals and the different mineral phases. In other words, the mineral materials have a higher specificity for metal sorption whereas the organic material would be less metal-selective, as shown by more similar Kd values for the five trace metals. Spatial VariabilitysMaps. The spatial variability in our data was evaluated, and a high degree of similarity is observed
TABLE 4. Empirical Regression Equations Resulting from a Multiple Stepwise Linear Regression Analysisa regression
N
R2
SEE
Kd (Cd) ) - 46Corg + 3488 Kd (Cu) ) 67Total Cu - 157ExMono + 364 Kd (Ni) ) 279TotalNi - 243ExMn 94ExMono - 69ExAl + 389 Kd (Pb) ) 68TotalPb - 3.66DOC + 1698 Kd (Zn) ) 371pH - 116ExMn 91ExMono - 739
54 59 52
0.369 0.205 0.766
398 373 237
12 60
0.916 0.552
470 150
a All regressions are significant at the 0.1% level, and only factors significant at the 5% level were included. The regression parameters are written in their order of increasing significance from left to right. The prefix Ex is used for exchangeable cations, Mono is used for monovalent ions (Na + K), Total is total recoverable metals using acid digestions (mg kg-1), Corg is the organic carbon content (% w/w), DOC is the dissolved organic carbon content (mg L-1), N is the sample size, and SEE is the standard error of estimate.
in the geographical distribution of total recoverable metals and of dissolved metals (5). For the Kd coefficients, the covariance with each of total recoverable metals, dissolved metals, or vegetation/species cover is much lower and thus allows the use of regression analysis. Evaluation of Controlling Factors. The results of the stepwise regression analyses to identify the most influential factors controlling metal partitioning in our data set are reported in Table 4. We expected soil pH and total metal contents to be controlling factors or at least highly significant properties. To our surprise, they were not significant in all cases. Also, the empirical regression equations reported in Table 4 have different structures, and they each include a different set of physicochemical predictors. No single factor is systematically present in all regressions. This lack of uniformity in the factors included in the regressions suggests that the regressions might be somewhat fortuitous and would hardly be applicable to another data set. The R2 values also show that some of the regression equations explain up to 92% of the variance while others explain only 21%. In all cases, the standard errors of estimates are quite high. Nevertheless, those regression equations may help to qualitatively identify which parameters are most influential in the determination of metal partitioning. It is noteworthy that each time an exchangeable cation factor is statistically significant, its coefficient is negative. The regression coefficients for exchangeable manganese, aluminum, iron and exchangeable monovalent cations are therefore all negative. This again suggests a circumstantial argument that these cations can actually compete for sorption and could possibly contribute to desorb trace metals. As expected, in the case where soil pH is significant, the coefficient is positive. Soil pH is an indicator of the protonation of exchanges sites on the solid phase. Similarly to exchangeable metals, increased proton competition means a higher release of trace metals to the soil solution. Nevertheless, the effect of pH is not straightforward since we also expect proton competition for complexation by dissolved organic ligands. That is, a lower pH will reduce the ability of dissolved organic substances to complex metals and reduce their ability to bring metals into the soil solution (which effectively contributes to increase Kd values), but at the same time, a lower pH will also reduce the solubility of organic matter. Table 4 and Figure 2 therefore suggest that, in the case of Zn, the positive effect of pH on its sorption to the reactive sites of the solid-phase organic matter is stronger than the effect on complexation by dissolved organic matter. On the other hand, the other regressions did not show a significant contribution of pH as a predictor. Given the undisputed importance of pH in controlling sorption reac-
tions, our data suggest that in this pH range, the effects of reduced metal sorption by the solid phase as H+ increases could actually be partly counteracted by a concomitant decrease in complexation with the dissolved ligands. Relations with Speciation Studies. Given the dichotomous effects of organic matter that can induce both sorption to the solid phase OM and complexation by dissolved organic matter in the solution, it is useful to compare our results with those focusing on relationships between solution free metal and sorbed (total) metal contents. We are specifically interested in comparing sorption to mineral versus organic materials. In studies on free metal activities in the soil solution, it is possible to discriminate the proportion of dissolved metal that is actually complexed with dissolved organic matter. McBride et al. (17) compared solution free Cu2+ activities following equilibration with either an iron oxide (R-goethite) or a peat soil. They noted that the organic material was characterized by free metal activities in solution that were 1 order of magnitude less those associated with iron oxide. Given that mineral soil horizons are only partly made of oxides, it is expected that a comparison of mineral and organic horizons would also demonstrate the stronger affinity of the organic portion of the soils to sorb Cu. In a similar experiment, a comparison was made of the solution free Pb2+ activities at equilibrium following sorption to leaf compost, iron oxides (synthetic ferrihydrite), and fieldcollected soil oxides (18). In contrast to the results for Cu2+, equilibrium solution Pb2+ activities were 2-6 orders of magnitude lower following sorption to the oxides relative to peat compost. But given that solution activities of free Pb2+ were still lower than those reported for Cu2+, this does not suggest that organic matter has a lower affinity for Pb relative to Cu. Instead it highlights the tremendous affinity of iron oxides for Pb. To illustrate this point, McBride et al. (17) have reported an equilibrium free Cu2+ of ∼10-8 M following sorption of 200 mg of Cu kg-1 to a peat soil at pH 6.5. Then, the Pb sorption study (18) reported an equilibrium free Pb2+ of ∼5 ×10-9 M following the addition 200 mg of Pb kg-1 and normalizing for a pH of 6.5. In summary, these studies are in agreement with the mean Kd values in Table 3, and they suggest that organic matter has a similar or higher affinity for Pb relative to Cu. Pb, however, has a much stronger affinity for the mineral component of soils. Table 4 also shows that dissolved organic matter is significant predictor of the Kd coefficient for only one metal, Pb. This reemphasizes that organic matter significantly affects Pb solubility (20) but that mineral solubility or Pb sorption to mineral phases dominates the behavior of Pb in soils. This has been illustrated with various publications showing that soil environments are conducive to the formation of chloropyromorphite and under some conditions, it actually controls Pb solubility (e.g., refs 13, 21, and 22). Risk Assessment Modeling. The differences between the Kd values of organic soils and of mineral soils are striking. Environmental fate and risk assessment modeling must therefore use the appropriate parameters. Yet, unlike the literature compilation for mineral soils, our data set is concentrated on a single small forested watershed. Comparison of our experimental Kd values with other data from organic soil horizons or peat soils will be necessary for a proper validation of Kd coefficients to parametrize environmental fate models. Nevertheless, the mean Kd values measured in this study should provide much better estimates of partitioning for acidic organic forest soil horizons than those compiled using mostly mineral soils.
Acknowledgments The authors are grateful for the financial support from the Metals in the Environment Research Network, the Universite´ VOL. 37, NO. 22, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5195
de Montre´al, the Toxic Substances Research Initiative, and the Natural Sciences and Engineering Research Council of Canada. The comments of three anonymous reviewers helped to improve the paper.
(10)
Supporting Information Available
(11) (12)
Figure 3 showing a digital elevation model representation of the Hermine Watershed. This material is available free of charge via the Internet at http://pubs.acs.org.
(13) (14)
Literature Cited (1) Sheppard, M. I.; Thibault, D. H. Health Phys. 1990, 59, 471-482. (2) Sauve´, S.; Hendershot, W.; Allen, H. E. Environ. Sci. Technol. 2000, 34, 1125-1131. (3) Soil Survey Staff. Keys to Soil Taxonomy, 4th ed.; SMSS Technical Monograph 19: Virginia Tech: Blacksburg, VA, 1990. (4) Sauve´, S. Speciation of metals in soils. In Bioavailability of Metals in Terrestrial Ecosystems: Importance of Partitioning for Bioavailability to Invertebrates, Microbes and Plants; Allen, H. E., Eds.; Society for Environmental Toxicology and Chemistry: Pensacola, FL, 2002; pp 7-58. (5) Manna, S.; Courchesne, F.; Roy, A. G.; Turmel, M.-C.; Cote´, B. Trace metals in the forest floor, litterfall patterns and topography. Transactions of the 17th World Congress of Soil Science, Bangkok, Thailand, 2002; pp 1306.1-1306.9. (6) Courchesne, F.; Roy, A. G.; Biron, P. M.; Coˆte´, B.; Fyles, J.; Hendershot, W. Environ. Monit. Assess. 2001, 67, 161-171. (7) Sheldrick, B. H. Analytical methods manual 1984; Land Resource Research Institute Contribution No. 84-30; Agriculture Canada: Ottawa, ON, Canada, 1984. (8) Hendershot, W. H.; Lalande, H.; Duquette, M. Ion exchange and exchangeable cations. In Soil Sampling and Methods of Analysis; Carter, M. R., Ed.; Lewis Publisher: Boca Raton, FL, 1993; pp 167-175. (9) Jones, J. B.; Case, V. W. Sampling, handling, and analyzing plant tissue samples. In Soil Testing and Plant Analysis, 3rd ed.; Book
5196
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 22, 2003
(15) (16) (17) (18) (19) (20) (21) (22) (23)
Series 3; Westerman, R. L., Ed.; Soil Science Society of America: Madison, WI, 1990; pp 390-427 Karam, A. Chemical properties of organic soils. In Soil Sampling and Methods of Analysis; Carter, M. R., Ed.; Lewis Publishers: Boca Raton, FL, 1993; pp 459-471. Chen, M.; Ma, L. Q. J. Environ. Qual. 1998, 27, 1294-1300. Temminghoff, E. J. M.; van der Zee, S. E. A. T. M.; de Haan, F. A. M. Eur. J. Soil Sci. 1998, 49, 617-628. Sauve´, S.; McBride, M.; Norvell, W. A.; Hendershot, W. Water, Air, Soil Pollut. 1997, 100, 133-149. Sauve´, S. The role of chemical speciation in bioavailability. In Bioavailability, Toxicity and Risk Relationships in Ecosystems; Naidu, R., Gupta, V. S. R., Kookana, R. S., Rogers, S., Adriano, D., Eds.; Science Publishers: Enfield, NH, 2003; pp 21-44. Buerge-Weirich, D.; Hari, R.; Xue, H.; Behra, P.; Sigg, L. Environ. Sci. Technol. 2002, 36, 328-336. Saar, R. A.; Weber, J. H. Geochim. Cosmochim. Acta 1980, 44, 1381. McBride, M.; Martı´nez, C. E.; Sauve´, S. Soil Sci. Soc. Am. J. 1998, 62, 1542-1548. Sauve´, S.; Martı´nez, C. E.; McBride, M.; Hendershot, W. Soil Sci. Soc. Am. J. 2000, 64, 595-599. Keller, C.; Ve´dy, J.-C. J. Environ. Qual. 1994, 23, 987-994. Turpeinen, R.; Salminen, J.; Kairesalo, T. Environ. Sci. Technol. 2000, 34, 5152-5156. Zhang, P.; Ryan, J. A.; Bryndzia, L. T. Environ Sci Technol 1997, 31, 2673-2678. Jopony, M.; Young, S. D. Eur. J. Soil Sci. 1994, 45, 59-70. Drouin, D. Ge´ne´ration d’un mode`le nume´rique d′e´le´vation ade´quat pour la mode´lisation hydrologique d’un petit bassin versant. M.Sc. Thesis, Department of Geography, Universite´ de Montre´al, Montre´al, PQ, Canada, 2000,
Received for review May 2, 2003. Revised manuscript received September 3, 2003. Accepted September 10, 2003. ES030059G
